Publication list
Selected publications, with short summaries.
The complete list is on ADS.
First author publications
ARCHITECTS I: Subgrid physics and the simulated CGM.
Subgrid models for star formation and supernova feedback are degenerate: they are usually calibrated to match galaxy stellar masses, so many different recipes can produce the same galaxy. To test how much this matters for the CGM, we introduce ARCHITECTS, a cosmological zoom-in simulation suite of the same galaxy, run with different subgrid models (mechanical, mechanical plus thermal, and delayed cooling) and tuned to the same stellar mass. Delayed cooling is bursty and ejective, leaving a gas-rich and metal-rich CGM. Mechanical feedback acts preventively at the galaxy scale, producing a gas-rich but metal-poor CGM. The hybrid model acts preventively at the halo scale, leaving a gas-poor CGM with metals concentrated around satellites. Galaxy properties alone cannot distinguish these models, but the CGM can, making it a key constraint on how stellar feedback regulates galaxy growth.
Subgrid models for star formation and supernova feedback are degenerate: they are usually calibrated to match galaxy stellar masses, so many different recipes can produce the same galaxy. To test how much this matters for the CGM, we introduce ARCHITECTS, a cosmological zoom-in simulation suite of the same galaxy, run with different subgrid models (mechanical, mechanical plus thermal, and delayed cooling) and tuned to the same stellar mass. Delayed cooling is bursty and ejective, leaving a gas-rich and metal-rich CGM. Mechanical feedback acts preventively at the galaxy scale, producing a gas-rich but metal-poor CGM. The hybrid model acts preventively at the halo scale, leaving a gas-poor CGM with metals concentrated around satellites. Galaxy properties alone cannot distinguish these models, but the CGM can, making it a key constraint on how stellar feedback regulates galaxy growth.
ARCHITECTS I: Subgrid physics and the simulated CGM.
Subgrid models for star formation and supernova feedback are degenerate: they are usually calibrated to match galaxy stellar masses, so many different recipes can produce the same galaxy. To test how much this matters for the CGM, we introduce ARCHITECTS, a cosmological zoom-in simulation suite of the same galaxy, run with different subgrid models (mechanical, mechanical plus thermal, and delayed cooling) and tuned to the same stellar mass. Delayed cooling is bursty and ejective, leaving a gas-rich and metal-rich CGM. Mechanical feedback acts preventively at the galaxy scale, producing a gas-rich but metal-poor CGM. The hybrid model acts preventively at the halo scale, leaving a gas-poor CGM with metals concentrated around satellites. Galaxy properties alone cannot distinguish these models, but the CGM can, making it a key constraint on how stellar feedback regulates galaxy growth.
ARCHITECTS I: Subgrid physics and the simulated CGM.
ARCHITECTS II: Subgrid physics and the observable CGM.
Subgrid models for star formation and supernova feedback are degenerate: they are usually calibrated to galaxy properties, so many different recipes can produce the same galaxy. The CGM, where feedback-driven outflows meet cosmological inflows, may break this degeneracy. We introduce ARCHITECTS, a cosmological zoom-in simulation suite of the same galaxy, run with subgrid models tuned to the same stellar mass. Delayed cooling is bursty and ejective, leaving a gas-rich and metal-rich CGM. Mechanical feedback acts preventively at the galaxy scale, leaving a gas-rich but low-metallicity CGM. The hybrid model acts preventively at the halo scale, creating a gas-poor CGM with metals concentrated around satellites. Galaxy properties alone cannot distinguish these models, but the CGM can, making it a key constraint on how stellar feedback regulates galaxy growth.
Subgrid models for star formation and supernova feedback are degenerate: they are usually calibrated to galaxy properties, so many different recipes can produce the same galaxy. The CGM, where feedback-driven outflows meet cosmological inflows, may break this degeneracy. We introduce ARCHITECTS, a cosmological zoom-in simulation suite of the same galaxy, run with subgrid models tuned to the same stellar mass. Delayed cooling is bursty and ejective, leaving a gas-rich and metal-rich CGM. Mechanical feedback acts preventively at the galaxy scale, leaving a gas-rich but low-metallicity CGM. The hybrid model acts preventively at the halo scale, creating a gas-poor CGM with metals concentrated around satellites. Galaxy properties alone cannot distinguish these models, but the CGM can, making it a key constraint on how stellar feedback regulates galaxy growth.
ARCHITECTS II: Subgrid physics and the observable CGM.
Subgrid models for star formation and supernova feedback are degenerate: they are usually calibrated to galaxy properties, so many different recipes can produce the same galaxy. The CGM, where feedback-driven outflows meet cosmological inflows, may break this degeneracy. We introduce ARCHITECTS, a cosmological zoom-in simulation suite of the same galaxy, run with subgrid models tuned to the same stellar mass. Delayed cooling is bursty and ejective, leaving a gas-rich and metal-rich CGM. Mechanical feedback acts preventively at the galaxy scale, leaving a gas-rich but low-metallicity CGM. The hybrid model acts preventively at the halo scale, creating a gas-poor CGM with metals concentrated around satellites. Galaxy properties alone cannot distinguish these models, but the CGM can, making it a key constraint on how stellar feedback regulates galaxy growth.
ARCHITECTS II: Subgrid physics and the observable CGM.
Co-authored publications
Mitigating overcooling with sink star formation
Galaxy simulations often suffer from overcooling: gas collapses too efficiently and forms too many stars. Comparing a standard Schmidt-type star formation model with a sink-particle model, where stars grow by accreting gas, we find that sinks form stars in rapid bursts whose radiation disperses clumps before supernovae explode. Supernovae then go off in low-density gas and drive stronger outflows, which redistribute metals from the galaxy center out into the CGM. The result is ~3× less stellar mass, a ~10× higher Lyman continuum escape fraction, and better agreement with JWST metallicities and sizes.
Galaxy simulations often suffer from overcooling: gas collapses too efficiently and forms too many stars. Comparing a standard Schmidt-type star formation model with a sink-particle model, where stars grow by accreting gas, we find that sinks form stars in rapid bursts whose radiation disperses clumps before supernovae explode. Supernovae then go off in low-density gas and drive stronger outflows, which redistribute metals from the galaxy center out into the CGM. The result is ~3× less stellar mass, a ~10× higher Lyman continuum escape fraction, and better agreement with JWST metallicities and sizes.
Mitigating overcooling with sink star formation
Galaxy simulations often suffer from overcooling: gas collapses too efficiently and forms too many stars. Comparing a standard Schmidt-type star formation model with a sink-particle model, where stars grow by accreting gas, we find that sinks form stars in rapid bursts whose radiation disperses clumps before supernovae explode. Supernovae then go off in low-density gas and drive stronger outflows, which redistribute metals from the galaxy center out into the CGM. The result is ~3× less stellar mass, a ~10× higher Lyman continuum escape fraction, and better agreement with JWST metallicities and sizes.
Mitigating overcooling with sink star formation
Uncertainties in high-z SED fitting
JWST has revealed a large population of galaxies in the early Universe, but their stellar masses, star formation rates and metallicities are inferred by fitting spectral energy distributions, an approach that is hard to validate. We apply the SED-fitting code Bagpipes to mock JWST/NIRCam photometry of simulated z = 6 galaxies, where the true properties are known. Even for the intrinsic stellar light, with no dust or emission lines, stellar masses are overestimated by about 60% on average, and by up to a factor of five in low-mass galaxies with recent starbursts, while star formation rates are underestimated by a factor of two. The main culprits are age–metallicity degeneracies and poorly constrained star formation histories. Dust and nebular emission add further biases, which bands free of strong lines, such as F410M, help to reduce. What we infer about the first galaxies is only as reliable as the models used to fit them.
JWST has revealed a large population of galaxies in the early Universe, but their stellar masses, star formation rates and metallicities are inferred by fitting spectral energy distributions, an approach that is hard to validate. We apply the SED-fitting code Bagpipes to mock JWST/NIRCam photometry of simulated z = 6 galaxies, where the true properties are known. Even for the intrinsic stellar light, with no dust or emission lines, stellar masses are overestimated by about 60% on average, and by up to a factor of five in low-mass galaxies with recent starbursts, while star formation rates are underestimated by a factor of two. The main culprits are age–metallicity degeneracies and poorly constrained star formation histories. Dust and nebular emission add further biases, which bands free of strong lines, such as F410M, help to reduce. What we infer about the first galaxies is only as reliable as the models used to fit them.
Uncertainties in high-z SED fitting
JWST has revealed a large population of galaxies in the early Universe, but their stellar masses, star formation rates and metallicities are inferred by fitting spectral energy distributions, an approach that is hard to validate. We apply the SED-fitting code Bagpipes to mock JWST/NIRCam photometry of simulated z = 6 galaxies, where the true properties are known. Even for the intrinsic stellar light, with no dust or emission lines, stellar masses are overestimated by about 60% on average, and by up to a factor of five in low-mass galaxies with recent starbursts, while star formation rates are underestimated by a factor of two. The main culprits are age–metallicity degeneracies and poorly constrained star formation histories. Dust and nebular emission add further biases, which bands free of strong lines, such as F410M, help to reduce. What we infer about the first galaxies is only as reliable as the models used to fit them.
Uncertainties in high-z SED fitting
Jellyfish Galaxies in Magnetic Fields
Jellyfish galaxies, whose gas is stripped by ram pressure as they plunge through galaxy clusters, show how a galaxy’s ISM is fed into its CGM. We run radiation-magnetohydrodynamic simulations of a gas-rich dwarf galaxy, resolving its multiphase ISM, star formation and stellar feedback, and expose it to magnetized and unmagnetized intracluster winds. Magnetized winds strip the disk more efficiently and leave smooth, diffuse tails, whereas hydrodynamic winds produce clumpy ones; the galaxy’s own field has little effect. With magnetic fields, stripped gas barely mixes with the intracluster medium and forms stars only within ~20 kpc of the disk; without them, mixing produces warm clouds that cool and form stars 50-100 kpc downstream. The magnetization of the cluster gas, far more than that of the galaxy, thus sets how stripped ISM is deposited into the CGM: as a smooth, shielded tail near the galaxy, or as clouds that mix and cool far downstream.
Jellyfish galaxies, whose gas is stripped by ram pressure as they plunge through galaxy clusters, show how a galaxy’s ISM is fed into its CGM. We run radiation-magnetohydrodynamic simulations of a gas-rich dwarf galaxy, resolving its multiphase ISM, star formation and stellar feedback, and expose it to magnetized and unmagnetized intracluster winds. Magnetized winds strip the disk more efficiently and leave smooth, diffuse tails, whereas hydrodynamic winds produce clumpy ones; the galaxy’s own field has little effect. With magnetic fields, stripped gas barely mixes with the intracluster medium and forms stars only within ~20 kpc of the disk; without them, mixing produces warm clouds that cool and form stars 50-100 kpc downstream. The magnetization of the cluster gas, far more than that of the galaxy, thus sets how stripped ISM is deposited into the CGM: as a smooth, shielded tail near the galaxy, or as clouds that mix and cool far downstream.
Jellyfish Galaxies in Magnetic Fields
Jellyfish galaxies, whose gas is stripped by ram pressure as they plunge through galaxy clusters, show how a galaxy’s ISM is fed into its CGM. We run radiation-magnetohydrodynamic simulations of a gas-rich dwarf galaxy, resolving its multiphase ISM, star formation and stellar feedback, and expose it to magnetized and unmagnetized intracluster winds. Magnetized winds strip the disk more efficiently and leave smooth, diffuse tails, whereas hydrodynamic winds produce clumpy ones; the galaxy’s own field has little effect. With magnetic fields, stripped gas barely mixes with the intracluster medium and forms stars only within ~20 kpc of the disk; without them, mixing produces warm clouds that cool and form stars 50-100 kpc downstream. The magnetization of the cluster gas, far more than that of the galaxy, thus sets how stripped ISM is deposited into the CGM: as a smooth, shielded tail near the galaxy, or as clouds that mix and cool far downstream.
Jellyfish Galaxies in Magnetic Fields
ARCHITECTS-CR: Cosmic rays and and the observable CGM.
Observations find abundant cold gas in the CGM, which cosmological simulations struggle to reproduce. Cosmic rays, a non-thermal source of pressure, are a promising way to change that. We re-simulate the ARCHITECTS simulations with no cosmic rays and with two cosmic-ray diffusion coefficients. At medium diffusion, cosmic rays remain confined to the galaxy. At high diffusion, they reach the CGM, lowering its temperature, enriching it with metals and driving hot outflows much farther out. This brings C IV and O VI closer to observations, but the gas remains too warm to reproduce the observed Mg II.
Observations find abundant cold gas in the CGM, which cosmological simulations struggle to reproduce. Cosmic rays, a non-thermal source of pressure, are a promising way to change that. We re-simulate the ARCHITECTS simulations with no cosmic rays and with two cosmic-ray diffusion coefficients. At medium diffusion, cosmic rays remain confined to the galaxy. At high diffusion, they reach the CGM, lowering its temperature, enriching it with metals and driving hot outflows much farther out. This brings C IV and O VI closer to observations, but the gas remains too warm to reproduce the observed Mg II.
ARCHITECTS-CR: Cosmic rays and and the observable CGM.
Observations find abundant cold gas in the CGM, which cosmological simulations struggle to reproduce. Cosmic rays, a non-thermal source of pressure, are a promising way to change that. We re-simulate the ARCHITECTS simulations with no cosmic rays and with two cosmic-ray diffusion coefficients. At medium diffusion, cosmic rays remain confined to the galaxy. At high diffusion, they reach the CGM, lowering its temperature, enriching it with metals and driving hot outflows much farther out. This brings C IV and O VI closer to observations, but the gas remains too warm to reproduce the observed Mg II.